电动车动力总成噪声品质粒子群-向量机预测模型  

Sound Quality Prediction of Electric Power Train Noise Based on Particle Swarm Optimization and Support Vector Machine

作　　者：方源[1,2] 章桐[1,2,3] 陈霏霏[1,2] 郭荣[1,2] 

机构地区：[1]同济大学新能源汽车工程中心,上海201804 [2]同济大学汽车学院,上海201804 [3]同济大学中德学院,上海201804

出　　处：《西安交通大学学报》2016年第1期41-46,共6页Journal of Xi'an Jiaotong University

基　　金：国家"863计划"资助项目(20U11AA11A265);国家自然科学基金资助项目(51205290);中央高校基本科研业务费专项资金资助项目(1700219118)

摘　　要：为了实现电动车动力总成噪声品质的预测,以某集中驱动式电动车为例,在考虑动力总成辐射噪声品质频域特性和已设立的敏感频带能量比这一客观评价参数的基础上进行了心理声学参数,即响度、尖锐度、粗糙度、抖动度、语音清晰度等与主观评价的相关性分析,由此建立了电动车动力总成噪声品质粒子群支持向量机预测模型,内容涉及采用支持向量机建立噪声品质预测模型、利用粒子群优化算法对向量基惩罚因子及核函数参数进行优化,最后验证了敏感频带能量比评价参数的有效性。研究结果表明:敏感频带能量比与主观评价相关度达到0.946,可以较好地反映主观感受;基于粒子群支持向量机的噪声品质预测模型的平均相对误差和最大相对误差分别为2.0%和6.7%,表明以敏感频带能量比作为输入特征的粒子群优化支持向量机模型,在电动车动力总成噪声品质的预测精度上优于基于遗传算法优化及网格搜索优化的预测模型。An electric power train is taken as a sample to predict its radiation noise quality.Studying frequency characteristics of sound quality and sensitive frequency-band energy ratio,a correlation analysis is conducted between subject evaluation and psychoacoustics parameters including loudness,sharpness,roughness,fluctuation and articulation index.Then a predicting model of sound quality of electric powertrain is established based on particle swarm optimization（PSO）and support vector machine（SVM）.After optimizing the penalty factor of SVM and parameters of kernel function by PSO,the effectiveness of Ris confirmed.The results indicate that subjective feeling can be reflected by sensitive frequency band energy ratio with correlation coefficient being 0.946.The absolute value and maximum value of the relative error are 2.0%and 6.7% respectively,which shows that the prediction accuracy of PSO-SVM is superior to those of the genetic algorithm method and grid search method.

关 键 词：电动车动力总成 噪声品质 粒子群优化 支持向量机 敏感频带能量比 

分 类 号：U463.2[机械工程—车辆工程]